Choice of development language is realistically more of a business decision than a technical one. With the proliferation of REST APIs, there are no longer really any technical hurdles to using any language.
Scalability can be an issue, but honestly business scalability is a problem for most businesses long before technical scalability is. Ruby and Python are awesome, but it can be hard to scale a company with one of those languages because it's hard to find good programmers who know Ruby or Python. It's much easier to find good Java or C# developers.
At this point, given all the work the Apache foundation has done to make Java awesome, I don't really know I would consider another language from a business standpoint. Sure, from a technical standpoint Java isn't that exciting; but it's easy to scale, easy to hire for and relatively standard across a number of platforms. Java code is also relatively easy to maintain since it forces you into certain design patterns.
There are some domains where this wouldn't hold true (and data science is definitely one of them -- you'll have an easier time hiring a machine learning expert who knows Python than Java) but that's really just technology. In a production environment, you'd eventually wrap your Python code in REST APIs and write all the ETL and interface code in Java.
This is probably one of the most excellent argument with respect to why we choose mainstream technology I've ever read.
You're on the money with Apache foundation (and other great quality tools/libraries/software written in Java). The ecosystem is just too huge to ignore. The other day we saw ElasticSearch at the top of HN when they announced they just released v1.0.0. I know tons of hi-tech companies are using it.
Meanwhile, there's no group or organization that invest in Ruby or Python the way Apache does to Java.
It's kinda sad. I really really wish (and want to see) similar thing happen to Ruby/Python but it's 2014 and sadly it hasn't happened. Outside Rails, I don't see anything else in the Ruby world that reach that level of adoption (maybe Jekyll or Octopress). Django seems to become minority. Plone seems to quiet down. Turbogears is probably dead. Flask is on the rise. Twisted doesn't seem to be the defacto solution.
Just that in general, Ruby or Python community seems to be locked on CRUD for web-app but nothing else.
Ruby and Python are awesome, but it can be hard to scale a company with one of those languages because it's hard to find good programmers who know Ruby or Python.
Good programmers usually become good python programmers quickly if it has become their day-to-day job. But maybe we have different levels in mind when we say "good".
While I would love to hire a bunch of elite hackers who are able to adapt to any language in a week, the reality is that most of those guys work for Google or Facebook and aren't interested in writing code for business apps (and probably wouldn't stick around anywhere early in their career for more than a year or so anyway). What you end up being able to hire are the B students from CS programs at good state schools (Texas, Michigan, etc.) They're not bad developers by any means, just not uber-hackers.
A lot of these guys are developers because it pays well and affords them a comfortable life for them and their (potentially future) family, not because they love technology. And there's nothing wrong with that.
Scalability can be an issue, but honestly business scalability is a problem for most businesses long before technical scalability is. Ruby and Python are awesome, but it can be hard to scale a company with one of those languages because it's hard to find good programmers who know Ruby or Python. It's much easier to find good Java or C# developers.
At this point, given all the work the Apache foundation has done to make Java awesome, I don't really know I would consider another language from a business standpoint. Sure, from a technical standpoint Java isn't that exciting; but it's easy to scale, easy to hire for and relatively standard across a number of platforms. Java code is also relatively easy to maintain since it forces you into certain design patterns.
There are some domains where this wouldn't hold true (and data science is definitely one of them -- you'll have an easier time hiring a machine learning expert who knows Python than Java) but that's really just technology. In a production environment, you'd eventually wrap your Python code in REST APIs and write all the ETL and interface code in Java.